PCI: add PCI Express ASPM support

PCI Express ASPM defines a protocol for PCI Express components in the D0
state to reduce Link power by placing their Links into a low power state
and instructing the other end of the Link to do likewise. This
capability allows hardware-autonomous, dynamic Link power reduction
beyond what is achievable by software-only controlled power management.
However, The device should be configured by software appropriately.
Enabling ASPM will save power, but will introduce device latency.

This patch adds ASPM support in Linux. It introduces a global policy for
ASPM, a sysfs file /sys/module/pcie_aspm/parameters/policy can control
it. The interface can be used as a boot option too. Currently we have
below setting:
        -default, BIOS default setting
        -powersave, highest power saving mode, enable all available ASPM
state and clock power management
        -performance, highest performance, disable ASPM and clock power
management
By default, the 'default' policy is used currently.

In my test, power difference between powersave mode and performance mode
is about 1.3w in a system with 3 PCIE links.

Note: some devices might not work well with aspm, either because chipset
issue or device issue. The patch provide API (pci_disable_link_state),
driver can disable ASPM for specific device.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

3 files changed, 834 insertions, 0 deletions

diff --git a/drivers/pci/pcie/Kconfig b/drivers/pci/pcie/Kconfig
index 287a9311716c..25b04fb2517d 100644
--- a/drivers/pci/pcie/Kconfig
+++ b/drivers/pci/pcie/Kconfig
@@ -26,3 +26,23 @@ config HOTPLUG_PCI_PCIE
 	  When in doubt, say N.
 
 source "drivers/pci/pcie/aer/Kconfig"
+
+#
+# PCI Express ASPM
+#
+config PCIEASPM
+	bool "PCI Express ASPM support(Experimental)"
+	depends on PCI && EXPERIMENTAL && PCIEPORTBUS
+	default y
+	help
+	  This enables PCI Express ASPM (Active State Power Management) and
+	  Clock Power Management. ASPM supports state L0/L0s/L1.
+
+	  When in doubt, say N.
+config PCIEASPM_DEBUG
+	bool "Debug PCI Express ASPM"
+	depends on PCIEASPM
+	default n
+	help
+	  This enables PCI Express ASPM debug support. It will add per-device
+	  interface to control ASPM.
diff --git a/drivers/pci/pcie/Makefile b/drivers/pci/pcie/Makefile
index e00fb99acf44..11f6bb1eae24 100644
--- a/drivers/pci/pcie/Makefile
+++ b/drivers/pci/pcie/Makefile
@@ -2,6 +2,9 @@
 # Makefile for PCI-Express PORT Driver
 #
 
+# Build PCI Express ASPM if needed
+obj-$(CONFIG_PCIEASPM)		+= aspm.o
+
 pcieportdrv-y			:= portdrv_core.o portdrv_pci.o portdrv_bus.o
 
 obj-$(CONFIG_PCIEPORTBUS)	+= pcieportdrv.o
diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c
new file mode 100644
index 000000000000..61fedb2448b6
--- /dev/null
+++ b/drivers/pci/pcie/aspm.c
@@ -0,0 +1,811 @@
+/*
+ * File:	drivers/pci/pcie/aspm.c
+ * Enabling PCIE link L0s/L1 state and Clock Power Management
+ *
+ * Copyright (C) 2007 Intel
+ * Copyright (C) Zhang Yanmin (yanmin.zhang@intel.com)
+ * Copyright (C) Shaohua Li (shaohua.li@intel.com)
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/pci.h>
+#include <linux/pci_regs.h>
+#include <linux/errno.h>
+#include <linux/pm.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/pci-aspm.h>
+#include "../pci.h"
+
+#ifdef MODULE_PARAM_PREFIX
+#undef MODULE_PARAM_PREFIX
+#endif
+#define MODULE_PARAM_PREFIX "pcie_aspm."
+
+struct endpoint_state {
+	unsigned int l0s_acceptable_latency;
+	unsigned int l1_acceptable_latency;
+};
+
+struct pcie_link_state {
+	struct list_head sibiling;
+	struct pci_dev *pdev;
+
+	/* ASPM state */
+	unsigned int support_state;
+	unsigned int enabled_state;
+	unsigned int bios_aspm_state;
+	/* upstream component */
+	unsigned int l0s_upper_latency;
+	unsigned int l1_upper_latency;
+	/* downstream component */
+	unsigned int l0s_down_latency;
+	unsigned int l1_down_latency;
+	/* Clock PM state*/
+	unsigned int clk_pm_capable;
+	unsigned int clk_pm_enabled;
+	unsigned int bios_clk_state;
+
+	/*
+	 * A pcie downstream port only has one slot under it, so at most there
+	 * are 8 functions
+	 */
+	struct endpoint_state endpoints[8];
+};
+
+static int aspm_disabled;
+static DEFINE_MUTEX(aspm_lock);
+static LIST_HEAD(link_list);
+
+#define POLICY_DEFAULT 0	/* BIOS default setting */
+#define POLICY_PERFORMANCE 1	/* high performance */
+#define POLICY_POWERSAVE 2	/* high power saving */
+static int aspm_policy;
+static const char *policy_str[] = {
+	[POLICY_DEFAULT] = "default",
+	[POLICY_PERFORMANCE] = "performance",
+	[POLICY_POWERSAVE] = "powersave"
+};
+
+static int policy_to_aspm_state(struct pci_dev *pdev)
+{
+	struct pcie_link_state *link_state = pdev->link_state;
+
+	switch (aspm_policy) {
+	case POLICY_PERFORMANCE:
+		/* Disable ASPM and Clock PM */
+		return 0;
+	case POLICY_POWERSAVE:
+		/* Enable ASPM L0s/L1 */
+		return PCIE_LINK_STATE_L0S|PCIE_LINK_STATE_L1;
+	case POLICY_DEFAULT:
+		return link_state->bios_aspm_state;
+	}
+	return 0;
+}
+
+static int policy_to_clkpm_state(struct pci_dev *pdev)
+{
+	struct pcie_link_state *link_state = pdev->link_state;
+
+	switch (aspm_policy) {
+	case POLICY_PERFORMANCE:
+		/* Disable ASPM and Clock PM */
+		return 0;
+	case POLICY_POWERSAVE:
+		/* Disable Clock PM */
+		return 1;
+	case POLICY_DEFAULT:
+		return link_state->bios_clk_state;
+	}
+	return 0;
+}
+
+static void pcie_set_clock_pm(struct pci_dev *pdev, int enable)
+{
+	struct pci_dev *child_dev;
+	int pos;
+	u16 reg16;
+	struct pcie_link_state *link_state = pdev->link_state;
+
+	list_for_each_entry(child_dev, &pdev->subordinate->devices, bus_list) {
+		pos = pci_find_capability(child_dev, PCI_CAP_ID_EXP);
+		if (!pos)
+			return;
+		pci_read_config_word(child_dev, pos + PCI_EXP_LNKCTL, &reg16);
+		if (enable)
+			reg16 |= PCI_EXP_LNKCTL_CLKREQ_EN;
+		else
+			reg16 &= ~PCI_EXP_LNKCTL_CLKREQ_EN;
+		pci_write_config_word(child_dev, pos + PCI_EXP_LNKCTL, reg16);
+	}
+	link_state->clk_pm_enabled = !!enable;
+}
+
+static void pcie_check_clock_pm(struct pci_dev *pdev)
+{
+	int pos;
+	u32 reg32;
+	u16 reg16;
+	int capable = 1, enabled = 1;
+	struct pci_dev *child_dev;
+	struct pcie_link_state *link_state = pdev->link_state;
+
+	/* All functions should have the same cap and state, take the worst */
+	list_for_each_entry(child_dev, &pdev->subordinate->devices, bus_list) {
+		pos = pci_find_capability(child_dev, PCI_CAP_ID_EXP);
+		if (!pos)
+			return;
+		pci_read_config_dword(child_dev, pos + PCI_EXP_LNKCAP, &reg32);
+		if (!(reg32 & PCI_EXP_LNKCAP_CLKPM)) {
+			capable = 0;
+			enabled = 0;
+			break;
+		}
+		pci_read_config_word(child_dev, pos + PCI_EXP_LNKCTL, &reg16);
+		if (!(reg16 & PCI_EXP_LNKCTL_CLKREQ_EN))
+			enabled = 0;
+	}
+	link_state->clk_pm_capable = capable;
+	link_state->clk_pm_enabled = enabled;
+	link_state->bios_clk_state = enabled;
+	pcie_set_clock_pm(pdev, policy_to_clkpm_state(pdev));
+}
+
+/*
+ * pcie_aspm_configure_common_clock: check if the 2 ends of a link
+ *   could use common clock. If they are, configure them to use the
+ *   common clock. That will reduce the ASPM state exit latency.
+ */
+static void pcie_aspm_configure_common_clock(struct pci_dev *pdev)
+{
+	int pos, child_pos;
+	u16 reg16 = 0;
+	struct pci_dev *child_dev;
+	int same_clock = 1;
+
+	/*
+	 * all functions of a slot should have the same Slot Clock
+	 * Configuration, so just check one function
+	 * */
+	child_dev = list_entry(pdev->subordinate->devices.next, struct pci_dev,
+		bus_list);
+	BUG_ON(!child_dev->is_pcie);
+
+	/* Check downstream component if bit Slot Clock Configuration is 1 */
+	child_pos = pci_find_capability(child_dev, PCI_CAP_ID_EXP);
+	pci_read_config_word(child_dev, child_pos + PCI_EXP_LNKSTA, &reg16);
+	if (!(reg16 & PCI_EXP_LNKSTA_SLC))
+		same_clock = 0;
+
+	/* Check upstream component if bit Slot Clock Configuration is 1 */
+	pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
+	pci_read_config_word(pdev, pos + PCI_EXP_LNKSTA, &reg16);
+	if (!(reg16 & PCI_EXP_LNKSTA_SLC))
+		same_clock = 0;
+
+	/* Configure downstream component, all functions */
+	list_for_each_entry(child_dev, &pdev->subordinate->devices, bus_list) {
+		child_pos = pci_find_capability(child_dev, PCI_CAP_ID_EXP);
+		pci_read_config_word(child_dev, child_pos + PCI_EXP_LNKCTL,
+			&reg16);
+		if (same_clock)
+			reg16 |= PCI_EXP_LNKCTL_CCC;
+		else
+			reg16 &= ~PCI_EXP_LNKCTL_CCC;
+		pci_write_config_word(child_dev, child_pos + PCI_EXP_LNKCTL,
+			reg16);
+	}
+
+	/* Configure upstream component */
+	pci_read_config_word(pdev, pos + PCI_EXP_LNKCTL, &reg16);
+	if (same_clock)
+		reg16 |= PCI_EXP_LNKCTL_CCC;
+	else
+		reg16 &= ~PCI_EXP_LNKCTL_CCC;
+	pci_write_config_word(pdev, pos + PCI_EXP_LNKCTL, reg16);
+
+	/* retrain link */
+	reg16 |= PCI_EXP_LNKCTL_RL;
+	pci_write_config_word(pdev, pos + PCI_EXP_LNKCTL, reg16);
+
+	/* Wait for link training end */
+	while (1) {
+		pci_read_config_word(pdev, pos + PCI_EXP_LNKSTA, &reg16);
+		if (!(reg16 & PCI_EXP_LNKSTA_LT))
+			break;
+		cpu_relax();
+	}
+}
+
+/*
+ * calc_L0S_latency: Convert L0s latency encoding to ns
+ */
+static unsigned int calc_L0S_latency(unsigned int latency_encoding, int ac)
+{
+	unsigned int ns = 64;
+
+	if (latency_encoding == 0x7) {
+		if (ac)
+			ns = -1U;
+		else
+			ns = 5*1000; /* > 4us */
+	} else
+		ns *= (1 << latency_encoding);
+	return ns;
+}
+
+/*
+ * calc_L1_latency: Convert L1 latency encoding to ns
+ */
+static unsigned int calc_L1_latency(unsigned int latency_encoding, int ac)
+{
+	unsigned int ns = 1000;
+
+	if (latency_encoding == 0x7) {
+		if (ac)
+			ns = -1U;
+		else
+			ns = 65*1000; /* > 64us */
+	} else
+		ns *= (1 << latency_encoding);
+	return ns;
+}
+
+static void pcie_aspm_get_cap_device(struct pci_dev *pdev, u32 *state,
+	unsigned int *l0s, unsigned int *l1, unsigned int *enabled)
+{
+	int pos;
+	u16 reg16;
+	u32 reg32;
+	unsigned int latency;
+
+	pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
+	pci_read_config_dword(pdev, pos + PCI_EXP_LNKCAP, &reg32);
+	*state = (reg32 & PCI_EXP_LNKCAP_ASPMS) >> 10;
+	if (*state != PCIE_LINK_STATE_L0S &&
+		*state != (PCIE_LINK_STATE_L1|PCIE_LINK_STATE_L0S))
+		*state = 0;
+	if (*state == 0)
+		return;
+
+	latency = (reg32 & PCI_EXP_LNKCAP_L0SEL) >> 12;
+	*l0s = calc_L0S_latency(latency, 0);
+	if (*state & PCIE_LINK_STATE_L1) {
+		latency = (reg32 & PCI_EXP_LNKCAP_L1EL) >> 15;
+		*l1 = calc_L1_latency(latency, 0);
+	}
+	pci_read_config_word(pdev, pos + PCI_EXP_LNKCTL, &reg16);
+	*enabled = reg16 & (PCIE_LINK_STATE_L0S|PCIE_LINK_STATE_L1);
+}
+
+static void pcie_aspm_cap_init(struct pci_dev *pdev)
+{
+	struct pci_dev *child_dev;
+	u32 state, tmp;
+	struct pcie_link_state *link_state = pdev->link_state;
+
+	/* upstream component states */
+	pcie_aspm_get_cap_device(pdev, &link_state->support_state,
+		&link_state->l0s_upper_latency,
+		&link_state->l1_upper_latency,
+		&link_state->enabled_state);
+	/* downstream component states, all functions have the same setting */
+	child_dev = list_entry(pdev->subordinate->devices.next, struct pci_dev,
+		bus_list);
+	pcie_aspm_get_cap_device(child_dev, &state,
+		&link_state->l0s_down_latency,
+		&link_state->l1_down_latency,
+		&tmp);
+	link_state->support_state &= state;
+	if (!link_state->support_state)
+		return;
+	link_state->enabled_state &= link_state->support_state;
+	link_state->bios_aspm_state = link_state->enabled_state;
+
+	/* ENDPOINT states*/
+	list_for_each_entry(child_dev, &pdev->subordinate->devices, bus_list) {
+		int pos;
+		u32 reg32;
+		unsigned int latency;
+		struct endpoint_state *ep_state =
+			&link_state->endpoints[PCI_FUNC(child_dev->devfn)];
+
+		if (child_dev->pcie_type != PCI_EXP_TYPE_ENDPOINT &&
+			child_dev->pcie_type != PCI_EXP_TYPE_LEG_END)
+			continue;
+
+		pos = pci_find_capability(child_dev, PCI_CAP_ID_EXP);
+		pci_read_config_dword(child_dev, pos + PCI_EXP_DEVCAP, &reg32);
+		latency = (reg32 & PCI_EXP_DEVCAP_L0S) >> 6;
+		latency = calc_L0S_latency(latency, 1);
+		ep_state->l0s_acceptable_latency = latency;
+		if (link_state->support_state & PCIE_LINK_STATE_L1) {
+			latency = (reg32 & PCI_EXP_DEVCAP_L1) >> 9;
+			latency = calc_L1_latency(latency, 1);
+			ep_state->l1_acceptable_latency = latency;
+		}
+	}
+}
+
+static unsigned int __pcie_aspm_check_state_one(struct pci_dev *pdev,
+	unsigned int state)
+{
+	struct pci_dev *parent_dev, *tmp_dev;
+	unsigned int latency, l1_latency = 0;
+	struct pcie_link_state *link_state;
+	struct endpoint_state *ep_state;
+
+	parent_dev = pdev->bus->self;
+	link_state = parent_dev->link_state;
+	state &= link_state->support_state;
+	if (state == 0)
+		return 0;
+	ep_state = &link_state->endpoints[PCI_FUNC(pdev->devfn)];
+
+	/*
+	 * Check latency for endpoint device.
+	 * TBD: The latency from the endpoint to root complex vary per
+	 * switch's upstream link state above the device. Here we just do a
+	 * simple check which assumes all links above the device can be in L1
+	 * state, that is we just consider the worst case. If switch's upstream
+	 * link can't be put into L0S/L1, then our check is too strictly.
+	 */
+	tmp_dev = pdev;
+	while (state & (PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1)) {
+		parent_dev = tmp_dev->bus->self;
+		link_state = parent_dev->link_state;
+		if (state & PCIE_LINK_STATE_L0S) {
+			latency = max_t(unsigned int,
+					link_state->l0s_upper_latency,
+					link_state->l0s_down_latency);
+			if (latency > ep_state->l0s_acceptable_latency)
+				state &= ~PCIE_LINK_STATE_L0S;
+		}
+		if (state & PCIE_LINK_STATE_L1) {
+			latency = max_t(unsigned int,
+					link_state->l1_upper_latency,
+					link_state->l1_down_latency);
+			if (latency + l1_latency >
+					ep_state->l1_acceptable_latency)
+				state &= ~PCIE_LINK_STATE_L1;
+		}
+		if (!parent_dev->bus->self) /* parent_dev is a root port */
+			break;
+		else {
+			/*
+			 * parent_dev is the downstream port of a switch, make
+			 * tmp_dev the upstream port of the switch
+			 */
+			tmp_dev = parent_dev->bus->self;
+			/*
+			 * every switch on the path to root complex need 1 more
+			 * microsecond for L1. Spec doesn't mention L0S.
+			 */
+			if (state & PCIE_LINK_STATE_L1)
+				l1_latency += 1000;
+		}
+	}
+	return state;
+}
+
+static unsigned int pcie_aspm_check_state(struct pci_dev *pdev,
+	unsigned int state)
+{
+	struct pci_dev *child_dev;
+
+	/* If no child, disable the link */
+	if (list_empty(&pdev->subordinate->devices))
+		return 0;
+	list_for_each_entry(child_dev, &pdev->subordinate->devices, bus_list) {
+		if (child_dev->pcie_type == PCI_EXP_TYPE_PCI_BRIDGE) {
+			/*
+			 * If downstream component of a link is pci bridge, we
+			 * disable ASPM for now for the link
+			 * */
+			state = 0;
+			break;
+		}
+		if ((child_dev->pcie_type != PCI_EXP_TYPE_ENDPOINT &&
+			child_dev->pcie_type != PCI_EXP_TYPE_LEG_END))
+			continue;
+		/* Device not in D0 doesn't need check latency */
+		if (child_dev->current_state == PCI_D1 ||
+			child_dev->current_state == PCI_D2 ||
+			child_dev->current_state == PCI_D3hot ||
+			child_dev->current_state == PCI_D3cold)
+			continue;
+		state = __pcie_aspm_check_state_one(child_dev, state);
+	}
+	return state;
+}
+
+static void __pcie_aspm_config_one_dev(struct pci_dev *pdev, unsigned int state)
+{
+	u16 reg16;
+	int pos = pci_find_capability(pdev, PCI_CAP_ID_EXP);
+
+	pci_read_config_word(pdev, pos + PCI_EXP_LNKCTL, &reg16);
+	reg16 &= ~0x3;
+	reg16 |= state;
+	pci_write_config_word(pdev, pos + PCI_EXP_LNKCTL, reg16);
+}
+
+static void __pcie_aspm_config_link(struct pci_dev *pdev, unsigned int state)
+{
+	struct pci_dev *child_dev;
+	int valid = 1;
+	struct pcie_link_state *link_state = pdev->link_state;
+
+	/*
+	 * if the downstream component has pci bridge function, don't do ASPM
+	 * now
+	 */
+	list_for_each_entry(child_dev, &pdev->subordinate->devices, bus_list) {
+		if (child_dev->pcie_type == PCI_EXP_TYPE_PCI_BRIDGE) {
+			valid = 0;
+			break;
+		}
+	}
+	if (!valid)
+		return;
+
+	/*
+	 * spec 2.0 suggests all functions should be configured the same
+	 * setting for ASPM. Enabling ASPM L1 should be done in upstream
+	 * component first and then downstream, and vice versa for disabling
+	 * ASPM L1. Spec doesn't mention L0S.
+	 */
+	if (state & PCIE_LINK_STATE_L1)
+		__pcie_aspm_config_one_dev(pdev, state);
+
+	list_for_each_entry(child_dev, &pdev->subordinate->devices, bus_list)
+		__pcie_aspm_config_one_dev(child_dev, state);
+
+	if (!(state & PCIE_LINK_STATE_L1))
+		__pcie_aspm_config_one_dev(pdev, state);
+
+	link_state->enabled_state = state;
+}
+
+static void __pcie_aspm_configure_link_state(struct pci_dev *pdev,
+	unsigned int state)
+{
+	struct pcie_link_state *link_state = pdev->link_state;
+
+	if (link_state->support_state == 0)
+		return;
+	state &= PCIE_LINK_STATE_L0S|PCIE_LINK_STATE_L1;
+
+	/* state 0 means disabling aspm */
+	state = pcie_aspm_check_state(pdev, state);
+	if (link_state->enabled_state == state)
+		return;
+	__pcie_aspm_config_link(pdev, state);
+}
+
+/*
+ * pcie_aspm_configure_link_state: enable/disable PCI express link state
+ * @pdev: the root port or switch downstream port
+ */
+static void pcie_aspm_configure_link_state(struct pci_dev *pdev,
+	unsigned int state)
+{
+	down_read(&pci_bus_sem);
+	mutex_lock(&aspm_lock);
+	__pcie_aspm_configure_link_state(pdev, state);
+	mutex_unlock(&aspm_lock);
+	up_read(&pci_bus_sem);
+}
+
+static void free_link_state(struct pci_dev *pdev)
+{
+	kfree(pdev->link_state);
+	pdev->link_state = NULL;
+}
+
+/*
+ * pcie_aspm_init_link_state: Initiate PCI express link state.
+ * It is called after the pcie and its children devices are scaned.
+ * @pdev: the root port or switch downstream port
+ */
+void pcie_aspm_init_link_state(struct pci_dev *pdev)
+{
+	unsigned int state;
+	struct pcie_link_state *link_state;
+	int error = 0;
+
+	if (aspm_disabled || !pdev->is_pcie || pdev->link_state)
+		return;
+	if (pdev->pcie_type != PCI_EXP_TYPE_ROOT_PORT &&
+		pdev->pcie_type != PCI_EXP_TYPE_DOWNSTREAM)
+		return;
+	down_read(&pci_bus_sem);
+	if (list_empty(&pdev->subordinate->devices))
+		goto out;
+
+	mutex_lock(&aspm_lock);
+
+	link_state = kzalloc(sizeof(*link_state), GFP_KERNEL);
+	if (!link_state)
+		goto unlock_out;
+	pdev->link_state = link_state;
+
+	pcie_aspm_configure_common_clock(pdev);
+
+	pcie_aspm_cap_init(pdev);
+
+	/* config link state to avoid BIOS error */
+	state = pcie_aspm_check_state(pdev, policy_to_aspm_state(pdev));
+	__pcie_aspm_config_link(pdev, state);
+
+	pcie_check_clock_pm(pdev);
+
+	link_state->pdev = pdev;
+	list_add(&link_state->sibiling, &link_list);
+
+unlock_out:
+	if (error)
+		free_link_state(pdev);
+	mutex_unlock(&aspm_lock);
+out:
+	up_read(&pci_bus_sem);
+}
+
+/* @pdev: the endpoint device */
+void pcie_aspm_exit_link_state(struct pci_dev *pdev)
+{
+	struct pci_dev *parent = pdev->bus->self;
+	struct pcie_link_state *link_state = parent->link_state;
+
+	if (aspm_disabled || !pdev->is_pcie || !parent || !link_state)
+		return;
+	if (parent->pcie_type != PCI_EXP_TYPE_ROOT_PORT &&
+		parent->pcie_type != PCI_EXP_TYPE_DOWNSTREAM)
+		return;
+	down_read(&pci_bus_sem);
+	mutex_lock(&aspm_lock);
+
+	/*
+	 * All PCIe functions are in one slot, remove one function will remove
+	 * the the whole slot, so just wait
+	 */
+	if (!list_empty(&parent->subordinate->devices))
+		goto out;
+
+	/* All functions are removed, so just disable ASPM for the link */
+	__pcie_aspm_config_one_dev(parent, 0);
+	list_del(&link_state->sibiling);
+	/* Clock PM is for endpoint device */
+
+	free_link_state(parent);
+out:
+	mutex_unlock(&aspm_lock);
+	up_read(&pci_bus_sem);
+}
+
+/* @pdev: the root port or switch downstream port */
+void pcie_aspm_pm_state_change(struct pci_dev *pdev)
+{
+	struct pcie_link_state *link_state = pdev->link_state;
+
+	if (aspm_disabled || !pdev->is_pcie || !pdev->link_state)
+		return;
+	if (pdev->pcie_type != PCI_EXP_TYPE_ROOT_PORT &&
+		pdev->pcie_type != PCI_EXP_TYPE_DOWNSTREAM)
+		return;
+	/*
+	 * devices changed PM state, we should recheck if latency meets all
+	 * functions' requirement
+	 */
+	pcie_aspm_configure_link_state(pdev, link_state->enabled_state);
+}
+
+/*
+ * pci_disable_link_state - disable pci device's link state, so the link will
+ * never enter specific states
+ */
+void pci_disable_link_state(struct pci_dev *pdev, int state)
+{
+	struct pci_dev *parent = pdev->bus->self;
+	struct pcie_link_state *link_state;
+
+	if (aspm_disabled || !pdev->is_pcie)
+		return;
+	if (pdev->pcie_type == PCI_EXP_TYPE_ROOT_PORT ||
+	    pdev->pcie_type == PCI_EXP_TYPE_DOWNSTREAM)
+		parent = pdev;
+	if (!parent || !parent->link_state)
+		return;
+
+	down_read(&pci_bus_sem);
+	mutex_lock(&aspm_lock);
+	link_state = parent->link_state;
+	link_state->support_state &=
+		~(state & (PCIE_LINK_STATE_L0S|PCIE_LINK_STATE_L1));
+	if (state & PCIE_LINK_STATE_CLKPM)
+		link_state->clk_pm_capable = 0;
+
+	__pcie_aspm_configure_link_state(parent, link_state->enabled_state);
+	if (!link_state->clk_pm_capable && link_state->clk_pm_enabled)
+		pcie_set_clock_pm(parent, 0);
+	mutex_unlock(&aspm_lock);
+	up_read(&pci_bus_sem);
+}
+EXPORT_SYMBOL(pci_disable_link_state);
+
+static int pcie_aspm_set_policy(const char *val, struct kernel_param *kp)
+{
+	int i;
+	struct pci_dev *pdev;
+	struct pcie_link_state *link_state;
+
+	for (i = 0; i < ARRAY_SIZE(policy_str); i++)
+		if (!strncmp(val, policy_str[i], strlen(policy_str[i])))
+			break;
+	if (i >= ARRAY_SIZE(policy_str))
+		return -EINVAL;
+	if (i == aspm_policy)
+		return 0;
+
+	down_read(&pci_bus_sem);
+	mutex_lock(&aspm_lock);
+	aspm_policy = i;
+	list_for_each_entry(link_state, &link_list, sibiling) {
+		pdev = link_state->pdev;
+		__pcie_aspm_configure_link_state(pdev,
+			policy_to_aspm_state(pdev));
+		if (link_state->clk_pm_capable &&
+		    link_state->clk_pm_enabled != policy_to_clkpm_state(pdev))
+			pcie_set_clock_pm(pdev, policy_to_clkpm_state(pdev));
+
+	}
+	mutex_unlock(&aspm_lock);
+	up_read(&pci_bus_sem);
+	return 0;
+}
+
+static int pcie_aspm_get_policy(char *buffer, struct kernel_param *kp)
+{
+	int i, cnt = 0;
+	for (i = 0; i < ARRAY_SIZE(policy_str); i++)
+		if (i == aspm_policy)
+			cnt += sprintf(buffer + cnt, "[%s] ", policy_str[i]);
+		else
+			cnt += sprintf(buffer + cnt, "%s ", policy_str[i]);
+	return cnt;
+}
+
+module_param_call(policy, pcie_aspm_set_policy, pcie_aspm_get_policy,
+	NULL, 0644);
+
+#ifdef CONFIG_PCIEASPM_DEBUG
+static ssize_t link_state_show(struct device *dev,
+		struct device_attribute *attr,
+		char *buf)
+{
+	struct pci_dev *pci_device = to_pci_dev(dev);
+	struct pcie_link_state *link_state = pci_device->link_state;
+
+	return sprintf(buf, "%d\n", link_state->enabled_state);
+}
+
+static ssize_t link_state_store(struct device *dev,
+		struct device_attribute *attr,
+		const char *buf,
+		size_t n)
+{
+	struct pci_dev *pci_device = to_pci_dev(dev);
+	int state;
+
+	if (n < 1)
+		return -EINVAL;
+	state = buf[0]-'0';
+	if (state >= 0 && state <= 3) {
+		/* setup link aspm state */
+		pcie_aspm_configure_link_state(pci_device, state);
+		return n;
+	}
+
+	return -EINVAL;
+}
+
+static ssize_t clk_ctl_show(struct device *dev,
+		struct device_attribute *attr,
+		char *buf)
+{
+	struct pci_dev *pci_device = to_pci_dev(dev);
+	struct pcie_link_state *link_state = pci_device->link_state;
+
+	return sprintf(buf, "%d\n", link_state->clk_pm_enabled);
+}
+
+static ssize_t clk_ctl_store(struct device *dev,
+		struct device_attribute *attr,
+		const char *buf,
+		size_t n)
+{
+	struct pci_dev *pci_device = to_pci_dev(dev);
+	int state;
+
+	if (n < 1)
+		return -EINVAL;
+	state = buf[0]-'0';
+
+	down_read(&pci_bus_sem);
+	mutex_lock(&aspm_lock);
+	pcie_set_clock_pm(pci_device, !!state);
+	mutex_unlock(&aspm_lock);
+	up_read(&pci_bus_sem);
+
+	return n;
+}
+
+static DEVICE_ATTR(link_state, 0644, link_state_show, link_state_store);
+static DEVICE_ATTR(clk_ctl, 0644, clk_ctl_show, clk_ctl_store);
+
+static char power_group[] = "power";
+void pcie_aspm_create_sysfs_dev_files(struct pci_dev *pdev)
+{
+	struct pcie_link_state *link_state = pdev->link_state;
+
+	if (!pdev->is_pcie || (pdev->pcie_type != PCI_EXP_TYPE_ROOT_PORT &&
+		pdev->pcie_type != PCI_EXP_TYPE_DOWNSTREAM) || !link_state)
+		return;
+
+	if (link_state->support_state)
+		sysfs_add_file_to_group(&pdev->dev.kobj,
+			&dev_attr_link_state.attr, power_group);
+	if (link_state->clk_pm_capable)
+		sysfs_add_file_to_group(&pdev->dev.kobj,
+			&dev_attr_clk_ctl.attr, power_group);
+}
+
+void pcie_aspm_remove_sysfs_dev_files(struct pci_dev *pdev)
+{
+	struct pcie_link_state *link_state = pdev->link_state;
+
+	if (!pdev->is_pcie || (pdev->pcie_type != PCI_EXP_TYPE_ROOT_PORT &&
+		pdev->pcie_type != PCI_EXP_TYPE_DOWNSTREAM) || !link_state)
+		return;
+
+	if (link_state->support_state)
+		sysfs_remove_file_from_group(&pdev->dev.kobj,
+			&dev_attr_link_state.attr, power_group);
+	if (link_state->clk_pm_capable)
+		sysfs_remove_file_from_group(&pdev->dev.kobj,
+			&dev_attr_clk_ctl.attr, power_group);
+}
+#endif
+
+static int __init pcie_aspm_disable(char *str)
+{
+	aspm_disabled = 1;
+	return 1;
+}
+
+__setup("pcie_noaspm", pcie_aspm_disable);
+
+#ifdef CONFIG_ACPI
+#include <acpi/acpi_bus.h>
+#include <linux/pci-acpi.h>
+static void pcie_aspm_platform_init(void)
+{
+	pcie_osc_support_set(OSC_ACTIVE_STATE_PWR_SUPPORT|
+		OSC_CLOCK_PWR_CAPABILITY_SUPPORT);
+}
+#else
+static inline void pcie_aspm_platform_init(void) { }
+#endif
+
+static int __init pcie_aspm_init(void)
+{
+	if (aspm_disabled)
+		return 0;
+	pcie_aspm_platform_init();
+	return 0;
+}
+
+fs_initcall(pcie_aspm_init);